Early Mars Hydrology: Meridiani Playa Deposits and the Sedimentary Record of Arabia Terra

Early Mars hydrology: Meridiani playa deposits and the sedimentary record of Arabia Terra

The MIT Faculty has made this article openly available. Please share how this access benefits you. Your story matters.

Citation Andrews-Hanna, Jeffrey C. et al. “Early Mars Hydrology: Meridiani Playa Deposits and the Sedimentary Record of Arabia Terra.” Journal of Geophysical Research 115.E6 (2010): n. pag. Web. ©2010 American Geophysical Union.

As Published http://dx.doi.org/10.1029/2009je003485

Publisher American Geophysical Union

Version Final published version

Citable link http://hdl.handle.net/1721.1/74246

Terms of Use Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, E06002, doi:10.1029/2009JE003485, 2010 Click Here for Full Article

Early Mars hydrology: Meridiani playa deposits and the sedimentary record of Arabia Terra Jeffrey C. Andrews‐Hanna,1 Maria T. Zuber,2 Raymond E. Arvidson,3 and Sandra M. Wiseman3 Received 31 July 2009; revised 14 November 2009; accepted 5 January 2010; published 5 June 2010.

[1] The Meridiani Planum region of Mars has been identified as a region of past aqueous activity by a combination of orbital and in situ observations that revealed evidence for sulfate‐rich dirty evaporites formed in a playa setting. We investigate the hydrology and sedimentary record of this area using global and regional hydrological models in which groundwater flow is driven by a combination of precipitation, evaporation, and the surface topography. Groundwater evaporation results in evaporite precipitation and cementation of aeolian sediments, allowing the accumulation of deposits of substantial thickness, which then affect the subsequent patterns of groundwater flow. Hydrological activity is initially predicted to be isolated to the deepest craters and depressions, primarily within the Arabia Terra region surrounding Meridiani. As these depressions fill with sediments, the groundwater upwelling spreads laterally across broad regions of Arabia Terra, including Meridiani Planum, as well as regional topographic lows such as the northern lowlands and large impact basins. The model predictions are borne out by observations of large intracrater deposits, inverted valley networks, finely layered deposits, spectral evidence for hydrated sulfates, and pedestal craters that preserve the remnants of a much larger deposit that once covered much of Arabia Terra. The results suggest that the inferred playa at Meridiani was part of a regionally extensive zone of groundwater upwelling. This hydrological cycle requires that conditions in the late Noachian to early Hesperian must have been conducive to the existence of liquid water at the surface throughout much of the low latitudes of Mars. Citation: Andrews‐Hanna, J. C., M. T. Zuber, R. E. Arvidson, and S. M. Wiseman (2010), Early Mars hydrology: Meridiani playa deposits and the sedimentary record of Arabia Terra, J. Geophys. Res., 115, E06002, doi:10.1029/2009JE003485.

1. Introduction (MOC) revealed finely layered rocks of apparent sedimentary origin concentrated in this region of Mars [Malin and [2] The surface of early Mars, like that of the Earth today, Edgett, 2000]. was shaped by an active hydrological cycle involving liquid [3] Motivated by these observations, the Mars Explora- water at and below the surface. The nature of the early tion Rover Opportunity was sent to explore this area, and hydrological cycle is thus far best revealed at Meridiani provided the first clear in situ evidence for the persistent Planum, where a combination of orbital and in situ observa- presence of liquid water on the surface of early Mars. The tions have revealed abundant evidence for water at an rover observations revealed that the hematite is present as a unprecedented level of detail. Early observations by the lag deposit of concretions concentrated above an eroding Mars Global Surveyor (MGS) Thermal Emission Spec- sedimentary deposit. Outcrops of the deposit exhibit both trometer (TES) revealed this region of Mars to be miner- aeolian and fluvial sedimentary structures [Grotzinger et al., alogically unique as the primary concentration of hematite, a 2005], with the grains themselves composed of a mixture of mineral commonly associated with aqueous processes sulfate salts and weathered basalt. The primary evaporite [Christensen et al., 2000, 2001]. At the same time, high mineralogy of the grains and the secondary reworked sedi- resolution images from the MGS Mars Orbiter Camera mentary structures have been overprinted by tertiary episodes of groundwater‐mediated diagenesis that were 1Department of Geophysics and Center for Space Resources, Colorado responsible for the formation of the hematite concretions, School of Mines, Golden, Colorado, USA. the cementation and alteration of the deposits, and the for- 2Department of Earth, Atmospheric, and Planetary Sciences, mation and dissolution of mineral clasts to leave behind Massachusetts Institute of Technology, Cambridge, Massachusetts, USA. tabular voids [Grotzinger et al., 2005; McLennan et al., 3Department of Earth and Planetary Sciences, Washington University in St. Louis, St. Louis, Missouri, USA. 2005; Squyres and Knoll, 2005]. [4] The Meridiani deposits have been interpreted as the Copyright 2010 by the American Geophysical Union. remains of an ancient playa in which dirty evaporites formed 0148‐0227/10/2009JE003485

E06002 1of22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002 in interdune depressions, were subjected to intensive aeolian [7] The results demonstrate that the Meridiani playa and fluvial reworking, and were then diagenetically modi- deposits are the natural outcome of a precipitation‐ fied by a fluctuating water table [Grotzinger et al., 2005; evaporation‐driven hydrological cycle operating under McLennan et al., 2005; Squyres and Knoll, 2005; Arvidson warm but arid conditions on early Mars. The preserved et al., 2006a]. The formation of these deposits and their deposits in Meridiani are predicted to be the remnant of a accumulation to thicknesses approaching 1 km must have once larger deposit covering much of Arabia Terra, the been accompanied and driven by a regional rise in the water vestiges of which are preserved elsewhere in the form of table [Arvidson et al., 2006a]. These interpretations are intracrater deposits, pedestal craters, and other deposits. bolstered by the association of similar sulfate and hematite Importantly, the hydrological cycle that drives the ground- deposits with locations of known groundwater outflow in water upwelling in Meridiani requires surface temperatures the chaos regions and canyons at the source of the outflow across the low to midlatitudes to have been above freezing channels [Glotch and Rogers, 2007]. Both the specific for long periods of time, in order to facilitate both the pre- interpretations of the observed deposits in Meridiani and cipitation‐induced recharge of the low latitude aquifers and the mechanisms invoked to produce them have been the groundwater evaporation upon reaching the surface. challenged by several alternative views [Knauth et al., While the climate at the time was likely quite arid by ter- 2005; McCollom and Hynek, 2005; Niles and Michalski, restrial standards, the surface conditions would have been 2009]. Nevertheless, the full suite of sedimentological conducive to Earth‐like life. and mineralogical evidence is best explained by deposition [8] The hydrological model and its implementation are of the sediments within a playa. In this work, we focus described in section 2. Section 3 investigates the global primarily on understanding the processes responsible for hydrological cycle in the late Noachian to early Hesperian the inferred playa deposit, and only briefly address these and the temporal evolution of groundwater flow and sedi- alternative interpretations. mentary deposition. Section 4 focuses on the regional [5] Meridiani Planum itself is part of a larger high thermal hydrology of the Arabia Terra region. In section 5, the inertia, light toned unit that is likely of similar provenance, model predictions are compared with the observed sedi- referred to as the etched terrain [Hynek, 2004]. In addition to mentary record of Arabia Terra. Alternative hypotheses for the deposits in Meridiani, widespread layered deposits of the origin of the deposits are discussed in light of the results apparent sedimentary origin are seen throughout Arabia and interpretations of this work in section 6. Finally, the Terra. Originally identified as a mantling unit [Moore, implications for the early climate and history of water on 1990], higher resolution MOC images revealed the depos- Mars are discussed in section 7. its to consist of finely layered sedimentary rocks [Malin and [9] In a companion paper (J. C. Andrews‐Hanna, Early Edgett, 2000; Edgett and Malin, 2002]. Evidence for Mars hydrology: 2. Hydrologic, climatic, and geochemical cementation and induration of the deposits suggests that evolution in the Noachian and Hesperian epochs, manu- liquid water likely played a role in their deposition or script in preparation, 2010), we focus on the temporal modification [Fassett and Head, 2007]. Morphological evolution of the hydrological cycle from the Noachian similarity and direct continuity of these Arabia Terra deposits through the early Hesperian. That paper places the Meridiani with the Meridiani deposits suggests a common origin, playa deposits within the broader context of the geomorphic though the mineralogy of the deposits across much of the and mineralogical evidence for the evolving climate of early region is obscured by dust. Thus it appears that the Mars, focusing on the nature and cause of the phyllosilicate hydrological processes responsible for playa formation in to sulfate deposition transition in the late Noachian [Bibring Meridiani may have been active across much of the Arabia et al., 2006]. Terra region. [6] Given the knowledge of the local environment at the time of deposition of the sedimentary sequence observed in 2. Groundwater Model Description Meridiani, the task remains to decipher the broader processes responsible for this environment and the implications [10] The early Martian hydrological cycle was likely for the climatic and hydrologic evolution of Mars. In a driven by precipitation. While there has been much con- previous study, we used global hydrological models of a tention over how warm and wet early Mars was [e.g., ‐ precipitation‐evaporation‐ driven hydrological cycle to Squyres and Kasting, 1994], some manner of precipitation demonstrate that Meridiani Planum and the surrounding induced recharge would have been necessary to recharge the ‐ Arabia Terra region is a preferred locus for groundwater low latitude aquifers in order to allow continued valley upwelling, driven primarily by the unique topography of network formation [Carr and Malin, 2000; Craddock and Arabia Terra [Andrews‐Hanna et al., 2007]. We here use Howard, 2002]. The question of whether this precipitation models of the global and regional scale groundwater was in the form of liquid rain [Craddock and Howard, hydrology to further investigate the distribution and 2002] or snowfall that would then melt and percolate into behavior of water on Mars during the late Noachian and the aquifers [Carr and Head, 2003] is of secondary early Hesperian epochs. We build on the earlier study by importance to the resulting hydrological evolution. In either ‐ considering the hydrological evolution at a higher spatial case, the precipitation would have recharged the low latitude resolution, explicitly including the formation of evaporites aquifers and, combined with the effects of the topography, and evaporite‐cemented sedimentary deposits, and com- would have driven groundwater flow down the local and paring the specific model predictions with the observed regional hydraulic gradients. Under arid conditions, sedimentary record. groundwater would flow through the aquifers until reaching the surface, at which point it would evaporate into the

2of22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

the system in a low‐latitude precipitation belt, with the relative latitudinal precipitation rate following a half cosine distribution between ± 45° latitude. This precipitation distribution is in first‐order agreement with the observed distribution of valley networks. While the cause of the observed latitudinal limit on valley network formation is unknown, it is reasonable to suppose that the conditions that prevented high‐latitude valley network formation likely in- hibited aquifer recharge as well. Such conditions may entail either a lack of high latitude precipitation, or persistently cold high latitude temperatures that prevented the runoff and infiltration of precipitation. [13] Models of groundwater flow on early Mars are limited by our ability to reconstruct the ancient surface topography. Tharsis appears to be an ancient feature [Phillips et al., 2001], though it is unclear how much of Tharsis was in place at the time of sulfate deposition in the late Noachian to early Hesperian. Previous work considered the effect of Figure 1. Topography in meters of (a) Mars (b) and Arabia removing the Tharsis load and induced deformation from Terra showing the location of the Meridiani deposits from the present‐day topography and the hydrological effects of Hynek [2004] (outline). (c) An approximation of the thick- subsequent Tharsis formation [Andrews‐Hanna et al., ness of the deposits, calculated by interpolating the sur- 2007]. Prior to Tharsis formation, the lack of high eleva- rounding topography inward, was removed from the tion perched aquifers on Tharsis would have resulted in present‐day topography prior to hydrological modeling. lower hydraulic heads on the periphery of the rise. During its formation, Tharsis may have acted as a hydrological sink due to the downward flexure of the lithosphere and burial of atmosphere, only to be redistributed on the surface as pre- aquifers deep beneath the rise. After Tharsis formation, the cipitation and continue to cycle through the system. expulsion of water from deeply buried aquifers may have [11] We represent the groundwater hydrology using a acted as a hydrological source. However, that work found fully explicit finite difference model representing the equations that the effect of Tharsis on the hydrology at Meridiani of unconfined groundwater flow in a spherical shell grid: Planum and elsewhere outside of the rise was second order @h 1 1 @ @h 1 @ @h compared to the effects of the regional topography of the ¼ K d þ K d sin dichotomy, Arabia Terra, and the large impact basins @t n r2 sin2 @ @ r2 sin @ @ [Andrews‐Hanna et al., 2007]. In this work, for the sake of þ sð; ; tÞ ð1Þ simplicity, we do not remove Tharsis from the present‐day topography. [14] On a local scale, many of the craters on the surface where h is the hydraulic head (referenced to the geoid), K is the today likely postdate the period of sulfate deposition, as vertically averaged hydraulic conductivity (m/s), n is the aquifer evidenced by the presence of large craters both super- porosity at the water table, d is the spatially varying height of the imposed on and interbedded within the Meridiani deposits. water table above the base of the aquifer (which is assumed to The present‐day topography includes both pristine unfilled lie at a constant depth below the surface), s isasourceterm craters and craters in various states of infill and exhumation representing the influx from precipitation (m/s), r is the radius [Malin and Edgett, 2000; Edgett and Malin, 2002]. While of the planet, is the longitude, ’ is the colatitude, and t is time. the specific short‐wavelength topography of the surface The aquifer porosity and hydraulic conductivity are taken from today is not an exact representation of the topography of the megaregolith model of Hanna and Phillips [2005]. The Mars in the late Noachian to early Hesperian, the early hydraulic conductivity is assigned to the vertically averaged topography of Mars would not have differed dramatically in value between the local water table and the base of the aquifer, nature. Young craters on the surface today are similar in and thus varies in both time and space during the simulations. form to the young craters on the surface at the time of [12] Flow through the aquifer is driven by the gradients in hydrological activity, while buried craters today do not the hydraulic head, which are established as a result of affect the groundwater flow paths in the models. precipitation‐induced recharge in the low latitudes and [15] As the primary focus of this study is the origin of the evaporative loss of groundwater in regions where the water Meridiani Planum sulfate deposits, we have removed a table reaches the surface. Conditions on Mars in the late crude representation of the added topography of the deposits Noachian and early Hesperian are assumed to be arid, in by interpolating the topography inward from the boundary agreement with the low erosion rates [Golombek et al., of the etched terrain and contiguous plains units [Hynek et 2006], low drainage densities, and immature state of the al., 2002; Hynek, 2004] using an inverse distance to a fluvial landscape [Stepinski and Stepinski, 2005; Barnhart et power scheme (Figure 1). The subtracted layer thickness al., 2009]. Thus, groundwater is assumed to evaporate upon ranging from 0 to 450 m is conservative, as the Meridiani reaching the surface, thereby limiting the water table to deposits are estimated to have a maximum thickness of remain at or beneath the surface. The globally integrated ∼600 m [Hynek et al., 2002]. Outside of the etched terrain, evaporative loss of water in each time step is added back to

3of22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002 we initialize the models using the present‐day geoid‐ to early Hesperian hydrology and the predicted distribu- referenced topography from MOLA [Smith et al., 2001]. tion of evaporite deposits. [16] Global models were run at a spatial resolution of [19] For an initial mean water table depth of 600 m, the 2.5 degrees per pixel, a limitation imposed by the model run water table at hydrological equilibrium lies deep below the times. These models are at a higher resolution than the surface over most of the planet, including the majority of 5 degrees per pixel results presented earlier [Andrews‐ the midlatitude precipitation belt. The water table lies at or Hanna et al., 2007]. Higher resolution simulations were near the surface in a few isolated locations, including the also performed locally at 0.25 degrees per pixel over a interiors of the giant impact basins, areas north of the limited region, using the global model results as a boundary dichotomy boundary, the Valles Marineris canyon system, condition. Further details of the model implementation, and Meridiani Planum and portions of the surrounding benchmarking, and parameter sensitivity can be found in the Arabia Terra region (Figure 2a). However, despite the broad appendices, including the effects of laterally heterogeneous regions of shallow groundwater, the water table only inter- aquifers and of different spatial distributions of precipitation sects the surface in a number of scattered points, generally on the surface. corresponding to the deepest craters and other depressions in regions with a shallow water table (Figures 2a and 2b). The effect of these local topographic lows in the steady state 3. Global Model Results: Evaporites models partially masks the long‐wavelength regional dis- and Sedimentary Deposits tribution of groundwater upwelling found in the earlier lower resolution models [Andrews‐Hanna et al., 2007], 3.1. Steady State Models though some concentration of loci of groundwater upwelling [17] We first consider the hydrological evolution of Mars is observed in Arabia Terra. The regional groundwater during the late Noachian to early Hesperian using steady upwelling throughout Arabia Terra only becomes apparent state models that neglect the modification of the surface that in models that include active sedimentary deposition, as would be expected to result from an active hydrological cycle. shown in the next section. Previous work found that the precipitation‐evaporation‐ driven hydrological cycle operating under arid conditions led 3.2. Dynamic Models: Effects of Sedimentary to the emergence of Meridiani Planum as a region of sus- Deposition tained groundwater upwelling and evaporation [Andrews‐ [20] The simplistic steady state model above provided a Hanna et al., 2007]. We now look at the hydrology at a static snapshot of the inherently dynamic hydrological higher resolution of 2.5 degrees per pixel. evolution of the planet. As a result of the arid climate [18] The models are best characterized in terms of the assumed in the models, groundwater reaching the surface initial mean water table depth implemented at the beginning will evaporate and leave behind any dissolved solutes in the of the simulation, which serves as a proxy for the total water form of evaporite deposits and evaporite‐cemented sediments. inventory of the planet. We focus on models with an initial As aeolian material is transported across the surface, it will mean water table depth of 600 m, for which the predicted be infiltrated by the rising groundwater table and indurated distribution of groundwater upwelling best matches the into an erosion‐resistant deposit. The resulting deposits may observed distribution of evaporites in Meridiani Planum. range from pure evaporitic salts to weakly cemented sand Rather than being a simple source of uncertainty, the initial and dust. The isolated topographic lows that intersect the mean water table depth is a parameter that would have water table will fill with sediments, allowing the water table experienced real variations over time during the early evo- in the surrounding area to rise toward the surface. The result lution of Mars. The total water inventory available to the is a dynamic hydrologic system in which the patterns of hydrologic system would have been subject to both short groundwater flow and evaporation alter the surface topog- and long‐term forcing as a result of periodic changes in the raphy, which in turn affects the groundwater flow. surface temperature and volume of ice in the cryosphere [21] This infilling of local topographic lows can result in driven by obliquity and orbital oscillations [Laskar et al., one of two outcomes, depending on the hydrologic envi- 2004], and secular loss of water due to impact and solar ronment: either the depression will partially fill with sedi- wind erosion of the atmosphere [Brain and Jakosky, 1998]. ments until the water table ceases to rise and the flux of The hydrological evolution for different assumed total water groundwater to the surface terminates, or the depression will inventories, representing different stages in the hydrologic‐ completely fill with sediments and the rising water table will climatic evolution of Mars, is examined in the companion intersect the surface throughout the surrounding area. A paper (manuscript in preparation, 2010). In short, that work similar process to the first scenario may have occurred on a finds that the progression from valley network and phyllo- much smaller scale after the impact that formed Meteor silicate formation in the early to middle Noachian, to Crater in Arizona, in which the excavated crater penetrated evaporitic sulfate deposition in the late Noachian to early to below the paleo water table leading to ancient lake and Hesperian can be understood as a consequence of the effect playa deposits on the crater floor [Kring, 2007]. The that the changing global water inventory had on the evolv- sedimentary infilling of Martian craters is supported by ing hydrological cycle. The declining total water inventory observations of layered sedimentary deposits within many resulted in a decrease in precipitation rates, a deepening of craters [Malin and Edgett, 2000], as discussed further in the water table, and a focusing of groundwater upwelling section 5. By smoothing out the short‐wavelength topo- within a few key regions of the planet, including Mer- graphic lows, the earlier low‐resolution models approximated idiani Planum and the surrounding Arabia Terra region. the hydrology of the planet in a more evolved state, after In this paper, we focus exclusively on the late Noachian substantial sedimentary infilling of the topographic lows.

4of22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

Figure 2. Time evolution of the global hydrological model over 50 Myr. (a) The depth to the water table varies little from its initial state. (b) The groundwater upwelling and evaporation is initially restricted to the deepest craters and low points. (c, e, g, i, and k) Over the course of the 50 Myr simulation, the deposits thicken and spread out over the surface while (d, f, h, j, and l) the rate of groundwater upwelling decreases. The location of Meridiani Planum and the surrounding etched terrain are outlined in each plot.

[22] In order to better represent the hydrologic evolution though the salinity will be strongly affected by the host rock of Mars, we now incorporate deposition of evaporites and lithology and the duration and temperature of water‐rock cemented sediments into the models. The water reaching the interaction. The mineralogy of the Meridiani deposits sug- surface at Meridiani and elsewhere has flowed underground gests that the fluids at the surface possessed higher solute at depths of up to a kilometer for distances of up to concentrations and lower water activities than terrestrial thousands of kilometers, with flow timescales on the order seawater [Tosca et al., 2008], though these salinities may of millions of years. Thus, there is ample time for these reflect the fluid composition after substantial evaporitic fluids to equilibrate with the aquifer matrix. Comparable concentration. We assume that the deep groundwater has a ancient fluids from deep drill holes on the Earth can have salinity equal to 80% that of seawater upon reaching the salinities exceeding that of seawater [Moller et al., 1997], surface [e.g., Handford, 1991], and that the deposits incor-

5of22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

of groundwater upwelling changes dramatically over the course of the simulation, the long wavelength shape of the water table is largely unaffected, since the actual changes to the surface topography are small. [24] As found in the earlier study, Meridiani Planum is one of the few regions of presently exposed Noachian‐aged crust for which a sustained flux of groundwater upwelling and evaporation is predicted. Andrews‐Hanna et al. [2007] showed that the groundwater upwelling in Meridiani is driven predominantly by the regional slopes of the Arabia Terra region. For the simple case of steady one‐dimensional flow, conservation of water volume results in an inverse relationship between the topographic slope and height of the water table relative to the surface, leading to a relative rise in the water table accompanying a transition from steep to gentle slopes. Arabia Terra is a unique region of Mars, Figure 3. Evolution of the equatorial precipitation rate intermediate between the highlands and lowlands in crustal during a global simulation. The precipitation distribution thickness and topography, and separated from each by dis- follows a cosine function between ± 45° latitude, with the tinct breaks in slope along its northern and southern rate set such that globally integrated precipitation exactly boundaries (Figures 1a and 1b). The location of Meridiani balances the globally integrated outflow and evaporation Planum just below the break in slope leading down into from the aquifers. Arabia Terra from the highlands results in a rise of the water table to the surface. This effect allows for a sustained flux of porate 50% by volume of clastic sediments based on mini‐ groundwater to the surface at Meridiani, despite its location TES observations at Meridiani [Glotch and Bandfield, 2006] along a sloping surface rather than within a confined basin. and similar to other “dirty” evaporites in playa settings, such [25] Predicted evaporite deposits in the northern lowlands that evaporation of 100 m of groundwater produces 2 m of and large impact basins would likely have been buried evaporitic sediments. The hydrologic system evolves slowly beneath subsequent volcanic and sedimentary deposits. in a state of quasi‐equilibrium with the changing surface Evaporite deposits are also predicted within the Valles topography, so different fluid salinities or clastic fractions Marineris canyons and in the nearby chaos regions and can be represented by simply scaling the model timescale. outflow channels, though it is uncertain to what extent these Meteoric rainwater is assumed to have a negligible con- topographic lows had developed by the late Noachian to centration of dissolved solutes, so any direct meteoric con- early Hesperian. This result agrees well with the detection of tribution to the local evaporation flux does not result in sulfates both in the Valles Marineris layered deposits evaporite deposition. The added sedimentary deposits are [Griffes et al., 2007; Murchie et al., 2009] and in layered assumed to have identical hydraulic properties to the deposits within Aram and other chaos regions [Glotch and underlying aquifers. Christensen, 2005; Glotch and Rogers, 2007; Noe Dobrea [23] The models are first run without evaporite deposition et al., 2008], and with the common association of hema- for 100 Myr to attain hydrological equilibrium, and then tite with these deposits [Christensen et al., 2000; Glotch and for an additional 100 Myr during which the topography is Christensen, 2005]. Regional hydrological models have updated to reflect evaporite deposition. When evaporite been used to investigate the groundwater flux and sedi- deposition is explicitly included in the models, the isolated mentary filling of Candor Chasma [Murchie et al., 2009]. craters and depressions in which the groundwater upwelling The local depositional environments in different settings was initially concentrated rapidly fill with evaporites. After may have ranged between lake, playa, spring mound, and 10 Myr, many of the initial loci of groundwater evaporation soil cementing deposits, leading to the observed morpho- are deactivated due to sedimentary infilling (Figures 2c and logical diversity. However, we suggest that the underlying 2d). In a few locations, sedimentary infilling of the local causes behind these widely separated sulfate deposits in topographic lows is accompanied by regional groundwater differing depositional settings are the same: long‐wave- rise, leading to broad regions of more evenly distributed length deep groundwater flow induced by a precipitation‐ groundwater evaporation. The distribution of groundwater evaporation‐driven hydrological cycle interacting with the evaporation across the surface becomes dominated by the surface topography. long‐wavelength patterns of groundwater flow, similar to [26] Prior to evaporite deposition, the equatorial precipi- the low‐resolution hydrological models of Andrews‐Hanna tation rate in balance with the globally integrated evapora- et al. [2007]. After 50 Myr, large‐scale evaporite deposits tion rate in the model is 1.55 × 10−4 m/year, and the mean in excess of several hundred meters thickness are found in groundwater flux at Meridiani is 1.43 × 10−4 m/year. As several locations in the northern lowlands, within the large sedimentary deposition fills in the local topographic lows, impact basins, in the Valles Marineris canyons, and con- the mean path length of groundwater flow increases. This centrated in Meridiani Planum and the surrounding Arabia increases the residence time of water in the subsurface, Terra region (Figure 2k). The smoothing of the topographic thereby decreasing the rate of water cycling between the lows has a diffusive effect on the distribution of ground- aquifers and atmosphere, and decreasing the precipitation water upwelling, leading to broad regions of smoothly rate (Figure 3). After 50 Myr, the equatorial precipitation varying groundwater flux (Figure 2l). While the distribution rate is only 7.43 × 10−5 m/year, and the groundwater flux at

6of22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

Meridiani is 2.56 × 10−5 m/year. Although the precipitation resolution by embedding regional hydrological models rate at Meridiani exceeds the groundwater flux to the sur- within the results of the global models. This approach is face, the groundwater would have carried a high solute similar to that taken by Harrison and Grimm [2008] in concentration in comparison to the dilute meteoric water. studying the outflow channels. The hydraulic head from the While constraints on the deposition rate at Meridiani are global models is used as a constant head boundary condition lacking, we show in section 5 that the predicted deposition around the periphery of the region of interest. The resolution rate at Becquerel crater matches that inferred from a com- of the steady state global model was first refined from 2.5 to parison of the layer periodicity with the obliquity evolution 1.25, and finally to 0.25 degrees per pixel, using the coarser of Mars [Lewis et al., 2008]. The predicted slow migration resolution model results as the initial condition in the suc- of fluids through the crust at Meridiani is consistent with the cessively higher resolution models, allowing rapid equili- stagnant aquifer conditions inferred from the small size and bration at the higher resolutions. The results of the high near‐spherical shape of the hematite concretions observed resolution steady state global model are then used as the by the Opportunity rover [McLennan et al., 2005; Sefton‐ initial condition and boundary conditions of a local dynamic Nash and Catling, 2008]. model encompassing the region of interest, in order to track [27] The specific rates of precipitation and evaporation the history of groundwater flow and evaporite deposition early in the simulations are sensitive to the model resolution. over longer periods of time. The precipitation rate in the models balances the globally [30] Applying constant head boundary conditions to the integrated evaporation rate, and thus provides a measure of local model neglects the small changes in topography and the rate of water cycling through the hydrologic system. At hydraulic head that should occur due to sedimentation higher resolutions, shorter wavelength craters and depres- beyond the margins of the region, but this has little affect on sions are resolved in regions of shallow water table. As a the evolution away from the boundaries. The precipitation result, the mean path length of groundwater flow decreases, rates must be dictated by the condition of global conserva- and the net rate of water cycling and the precipitation rate tion of water volume, and will change as evaporite deposi- increase. The equatorial precipitation rate for global models tion alters the mean path length and net rate of groundwater with grid resolutions of 2.5 to 0.25 degrees per pixel ranges cycling. The local models cannot track global water con- from 1.55 × 10−4 to 4.0 × 10−3 m/year. However, the model servation, so we instead apply the time‐evolving precipita- resolution has little effect on the regional patterns of tion rates from the lower resolution dynamic global models. groundwater flow or the depth to the water table. As the This will underestimate the precipitation and evaporation high resolution models evolve, the sedimentary infilling of rates early in the simulation, but have little effect as the the topographic lows increases the flow path lengths, the system evolves, other than to increase the time required to precipitation rates decrease, and the results converge with generate deposits of a given thickness in these local models those of the lower resolution models. relative to the global models. [28] The low predicted precipitation rates are consistent [31] The evolution of the local height of the water table in with the lack of evidence for modification of the deposits by Meridiani Planum and the surrounding Arabia Terra region dilute meteoric water. Comparable mean precipitation rates is determined primarily by recharge within the highlands to are observed locally in hyperarid regions on the Earth, the south and evaporation in the lowlands to the north including portions of the Atacama Desert of Chile that [Andrews‐Hanna et al., 2007], thereby dictating that the receive less than 1 mm of rainfall per year on average with regional model extend beyond the northern and southern significant interannual variability [Houston, 2006]. The edges of Arabia Terra. We define a rectangular region of predicted rates of groundwater upwelling and midlatitude interest in latitude‐longitude space enclosing Arabia Terra precipitation should be considered to be long‐term averaged and including a portion of the adjacent highlands and low- rates. In reality, Mars would have experienced climatic and lands (Figure 4b). As in the global models, the Meridiani hydrological perturbations that would have modulated these hematite unit and etched terrain have been subtracted from average rates. The predicted low precipitation rates may the topography. We have similarly filled in the depressions translate into long periods of hyperaridity punctuated by of the large outflow channels in the western portion of the occasional climatic optima due to orbital forcing [Laskar et region, though these do not strongly affect the groundwater al., 2004], in which the climate was similar to arid or flow in the rest of the region. semiarid environments on the Earth [Barnhart et al., 2009]. [32] Early in the simulation, evaporite formation is Similarly, the continued impact bombardment of the surface restricted to scattered impact craters and other topographic would have resulted in repeated perturbations to the lows, which are now resolved down to a scale of 10s of km. hydrologic cycle on a local to regional scale, with a large However, even at this early stage it is clear that the region fresh impact crater lowering the water table and inhibiting encompassing Meridiani Planum exhibits a particularly groundwater upwelling over a broad region, until sufficient shallow water table (Figure 4a). As the system evolves, the sedimentary infilling of that crater allowed the regional deepest depressions are filled with sediments, allowing the water table to again intersect the surface. water table to rise toward the surface over much of Arabia Terra (Figures 4d, 4g, 4j, and 4m), leading to broadly dis- 4. Regional Models: Arabia Terra and Meridiani tributed sedimentary deposits (Figures 4e, 4h, 4k, and 4n) and large areas of nearly uniform groundwater upwelling Planum (Figures 4f, 4i, 4l, and 4o). Looking at the distribution of [29] The resolution of the global models is limited by the evaporites after 200 Myr (Figure 4n), we find two classes of computational time required. However, it is possible to deposits. Local deposits with thicknesses of up to 1 km are investigate the hydrology of particular regions at higher found within many of the closed depressions, including the

7of22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

Figure 4. (a) The depth to the water table, (b) topography (with the Meridiani etched terrain and major outflow channels removed), and (c) groundwater upwelling rate and distribution (with groundwater upwelling restricted to isolated model pixels at t = 0 Myr) are shown at the initiation of the local hydrological model of Arabia Terra. (d, g, j, and m) As the model evolves, the water table rises toward the surface throughout the region, (e, h, k, and n) the deposits thicken and spread over the surface, and (f, i, l, and o) the groundwater upwelling and evaporation rate spreads over the surface and decreases in magnitude. All model results are superimposed over MOLA shaded relief. Boxes in Figure 4n show the locations of Figures 9a and 9b.

8of22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002 large impact craters and the chaos regions in western Arabia that none of the direct morphological observations on the Terra. Extensive evaporite deposits are also found spreading surface in Meridiani require active precipitation, mainte- out over the intercrater plains, particularly in the vicinity of nance of the hydrological environment responsible for the Meridiani Planum, that are not contained within closed playa deposits requires continued low‐latitude precipitation topographic depressions. After 200 Myr, the deposits in on Mars as late as the early Hesperian. Meridiani have reached a maximum thickness of 850 m and mean thickness of 120 m, comparable to the thickness of the 5. Comparison With the Sedimentary Record deposits removed from the original topography (450 m and 180 m, respectively). of Arabia Terra [33] The fact that the observed deposits are more limited [36] The hydrological models highlight Arabia Terra as a in lateral extent, and yet thicker on average and with steeper region of enhanced groundwater upwelling early in Mars margins than predicted by the models, suggests that they are history, and predict that widespread evaporite‐cemented the end product of a combination of primary evaporite/playa sedimentary deposits should have formed across much of deposition and secondary erosion and redeposition, consis- this region. While direct evidence for playa evaporites from tent with the in situ observations by Opportunity. A key in situ observations exists only at one location in Meridiani example of this secondary modification of the original de- Planum, there is abundant evidence for sedimentation and posits is found in Miyamoto crater at the southwest edge of hydrological activity throughout Arabia Terra. the Meridiani deposits. Wiseman et al. [2008] showed that [37] One of the key predictions of the models is that the the ancient phyllosilicate‐bearing Noachian floor is exposed large craters and other topographic lows in Arabia Terra in the southwest corner of the crater, immediately adjacent should have been filled with evaporites and cemented to thick stacks of sulfates in the northeast portion of the sediments prior to playa development at Meridiani. MOLA crater. That study argued that the deposits once filled the topography reveals a number of large, mound‐like sedi- crater and were later eroded back to their present extent. mentary deposits within craters (Figure 5). These deposits The groundwater models predict that evaporite deposition exhibit thicknesses of up to 2 km. The deposits in Crom- should have been active throughout the crater floor, with melin crater (Figures 5b and 5i) exceed the elevation of the peak rates in the southwest portion of the crater. The steep crater rim, suggesting that the deposits at one time filled the margin of the deposits filling the northeast portion of the crater and covered the surrounding Noachian plains. This is crater cannot be explained by hydrological processes alone, supported by the observation of pedestal craters outside of since a water table controlled by diffusive flow will assume Crommelin that reach a similar peak elevation to the a more gradual profile. Erosion and redeposition of the intracrater deposit, which further suggests that the majority sulfates could have occurred either during or after the initial of the deposits in Crommelin lie stratigraphically below the deposition. deposits on the surrounding intercrater plains. Distinct [34] The model predictions are consistent with the topographic benches in the deposits within Henry crater observed continuity of layers over distances of 100s of km suggest layers with varying degrees of induration (Figures 5c within Meridiani and the low dip angles that generally and 5j), possibly modulated by perturbations to the climatic‐ conform to the regional topography [Hynek and Phillips, hydrologic system during deposition. That the deposits at 2008]. After the infill of the large craters in the region, one time filled the craters rather than having formed with the water table is predicted to follow a smoothed version their current mound morphologies is evidenced by the of the surface topography. As the system evolves, broad presence of a topographic bench along the wall of Becquerel regions of relatively uniform groundwater evaporation are crater at the same elevation as the intracrater mound predicted over much of Arabia Terra (Figure 4o). This (Figures 5d and 5k). would gradually build up large‐scale deposits with coherent [38] The presence of thick layered deposits within craters layering over long distances through either direct evaporite lacking any obvious fluvial input is strongly suggestive of a deposition or the cementation of aeolian deposits. Hynek groundwater source, and previous studies have interpreted and Phillips [2008] find layer dips clustered between these and other light‐toned layered deposits as large‐scale 0.05° and 1.0°, with dip azimuths between −15° and −60° spring deposits [Rossi et al., 2008]. Many craters within measured clockwise from north. Considering the predicted Arabia Terra exhibit varying states of fill and exhumation deposit surface (representing the primary bedding plains) [Malin and Edgett, 2000], suggesting that the observed within the Meridiani etched terrain, we find dip azimuths of extent of sedimentary deposits may belie their original dis- −46 ± 18° and dips of 0.035 ± 0.008°. While the predicted tribution. Some craters lacking deposits may have been dip azimuths agree well with the observations, the predicted filled and then fully exhumed, while others may postdate the layer dips are on the low end of the measured range. The main period of groundwater upwelling and evaporite somewhat greater observed layer dips may be a result of deposition. either postdepositional deformation or increased hydraulic [39] For the deposits within Becquerel Crater, high gradients resulting from spatial variations in either the resolution images and DEM’s have allowed a direct corre- aquifer properties or the distribution of precipitation. lation of the observed layer bundling with the predicted [35] In simulations in which the precipitation rate was obliquity cycle of Mars, leading to a calculated deposition arbitrarily terminated, the groundwater flux at Meridiani rate of 3 × 10−5 m/year [Lewis et al., 2008]. This deposition ceased within 5 Myr. This brief period of continued rate is in excellent agreement with the predictions of the groundwater discharge after the termination of the precipi- local model for Becquerel of a groundwater upwelling rate tation‐induced recharge is insufficient to generate the of 10−3 m/year and deposition rate of 2 × 10−5 m/year early observed thickness of the deposits. Thus, despite the fact in the simulation (Figure 4f). While the predicted deposition

9of22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

[40] The composition of the intracrater deposits is masked by dust across much of the region, as evidenced by the low thermal inertia derived from TES [Putzig et al., 2005]. However, the Arabia Terra region as a whole is an area of enhanced hydration, as observed by the gamma ray and neutron spectrometer aboard Mars Odyssey [Boynton et al., 2002; Feldman et al., 2004]. This broad, diffuse signature of enhanced hydration may reflect the presence of both the hydrated sulfate deposits and the redistributed erosional debris from the deposits. In some locations, spectral sig- natures indicative of the presence of hydrated sulfates are observed in CRISM observations of the intracrater deposits. A hydrated sulfate parameter map was constructed from CRISM multispectral (MSP) data over part of Arabia Terra (using the D2100, D2400, and D1900 band depth parameters from Wiseman et al. [2010], similar to those of Pelkey et al. [2007]). Hydrated sulfate identifications are found associated with a number of large intracrater deposits, including one in southern Arabia Terra ∼300 km from the edge of the Meridiani etched terrain (Figure 6). Absorption features near 1.9 and 2.4 mm indicate the presence of polyhydrated sulfates [Cloutis et al., 2006]. The strongest sulfate detections in this case are at the bottom of a partially exhumed deposit, which may be due to a greater sulfate concentration or difference in sulfate mineralogy lower in the deposits. This would be consistent with the model prediction of a decreasing groundwater influx as the craters fill with sediment, which would result in a higher evaporite to clastic ratio lower in the stratigraphic section. [41] A survey of large craters (>100 km diameter) across Mars reveals a distinct concentration of intracrater sedimentary deposits in Arabia Terra (Figure 7a). Similarly, in a study of open crater lakes, Fassett and Head [2008] used the size of the upstream drainage basins to evaluate the relative importance of runoff versus groundwater inflow. They found that the open lake basins inferred to have experienced a greater contribution from groundwater inflow were concentrated primarily in Arabia Terra (Figure 7b). This is Figure 5. MOLA topography of select large intracrater consistent with the predicted shallow water table and ele- deposits, including (a) Miyamoto, (b) Crommelin, (c) Henry, vated ratio of groundwater flux to precipitation predicted (d) Becquerel, (e) Vernal crater, (f) a crater near Tuscaloosa throughout the region. crater, and (g) craters near Capen which is crater on right. [42] The sulfate‐hematite deposits within Aram chaos The presence of pedestal craters (“P”) on the surrounding represent an interesting and unique case of an intracrater intercrater plains suggest that deposits may have once been deposit [Christensen et al., 2000; Gendrin et al., 2005; more laterally extensive. Topographic profiles of (h) Miya- Glotch and Christensen, 2005; Lichtenberg et al., 2010]. moto, (i) Crommelin, (j) Henry, and (k) Bequerel reveal Aram crater lies in the westernmost portion of Arabia Terra, the thickness of the deposits. Topographic steps in Henry and hosts a chaos region that contributed to the flow within (Figure 5j, arrows) suggest large‐scale internal layering. A Ares Vallis. Spectral data reveal a km thick sulfate and bench in the wall of Becquerel at the same elevation as hematite deposit that is stratigraphically above the chaos on the intracrater deposit (Figure 5k, arrows) suggest that the crater floor. This intracrater deposit postdates the Mer- the deposits at one time filled the entire crater. The box idiani deposit by up to 1 Gyr [Glotch and Christensen, in Figure 5g reveals the location of the CRISM parameter 2005], yet exhibits striking mineralogical and morphologi- map in Figure 6a. cal similarities [Glotch and Christensen, 2005; Lichtenberg et al., 2010]. While the relationship between the intracrater deposit and the chaos in Aram provide a strong link between rate is highly sensitive to the poorly constrained aquifer the Arabia Terra intracrater deposits and groundwater hydraulic conductivity and the assumed ratio between the upwelling, the timing of evaporite deposition presents a deposit thickness and the evaporated water column, the conundrum. Chaos formation and outflow flooding are close agreement of the predicted and observationally commonly thought to require cold climate conditions and a inferred rates lends strong support to the interpretation that thick cryosphere, while the playa hypothesis for the origin these intracrater deposits formed as a result of groundwater of the Meridiani deposits requires evaporating liquid water inflow. at or near the surface. The sequence in Aram could be in-

10 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

deposits would also have formed within Aram crater prior to chaos formation and outflow flooding. This original intracrater deposit would have been disrupted and reworked by the emerging flood waters, and ultimately carried downstream in the flood and/or redeposited in the crater lake above the chaos. [43] In addition to the sedimentary infilling of large craters, other topographic lows should have been preferentially filled with evaporite‐cemented materials. Fluvial valleys, by nature, follow local topographic lows. If the valley networks received any significant fraction of their flow from groundwater, as is commonly the case on Earth, they would represent locations in which the surface topography inter- sected the paleo water table during periods of active formation. After the transition to a more arid hydrologic‐climatic

Figure 6. CRISM hydrated sulfate parameter map over a large intracrater deposit (see Figure 5g for context), cre- ated from volcano scan‐corrected CRISM I/F multispectral data (a) R, D2100; G, 2400; B, D1900, similar to the spectral parameters of Pelkey et al. [2007] superimposed over a THEMIS NIR mosaic also showing the CRISM MSP coverage. In this color scheme, polyhydrated sulfates are represented as cyan and monohydrated sulfates as yellow (not identified). (b) The spectrum reveals absorption features at 1.9 and 2.4 mm, consistent with the identification of polyhydrated sulfates. Brighter tones in the THEMIS NIR base map indicate surface materials with a greater degree of induration or larger grain size. terpreted either as a result of a significant climatic swing to warm conditions in the late Hesperian after outflow channel formation or, more likely, as a result of local conditions within Aram that permitted a depositional environment similar to that in Meridiani despite the prevailing cold climate [Glotch and Christensen, 2005]. Hydrological models of outflow channel floods from chaos regions demonstrate that the late stage fluids originate from the deepest portions of the aquifer [Andrews‐Hanna and Phillips, 2007], suggesting the possibility of a flux of warm deep brines to the surface and a hydrothermally enhanced heat flux. If evap- Figure 7. Evidence of hydrological activity and deposition oration of these upwelling brines outpaced freezing, a playa‐ in Arabia Terra, including (a) the distribution of large intra- like deposit similar to that in Meridiani may result even crater deposits, (b) the concentration of open basin lakes under cold climate conditions. This process would have showing evidence for a substantial groundwater contribution been aided by the freezing point depression resulting from (from Fassett and Head [2008]), and (c) the concentration evaporitic concentration of the salts in the near‐surface of raised curvilinear features representing inverted valley fluids. The widespread abundance of older intracrater networks (valley networks represented in black from Carr deposits across Arabia Terra coupled with the predictions and Chuang [1997]; raised curvilinear features represented of the hydrological models argue that substantial sulfate by circles from Williams [2007]).

11 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002 regime in the late Noachian (manuscript in preparation, [46] These deposits have experienced significant erosion 2010), the valley networks in Arabia Terra would be preferred since their initial deposition, as evidenced by erosional locations for evaporite deposition. While Arabia Terra surfaces cutting across the bedding, craters in varying states exhibits a lower density of valley networks than the rest of the of fill and exhumation [Malin and Edgett, 2000], and ero- equatorial southern highlands [Carr and Chuang, 1997], sional outliers surrounding the Meridiani deposits [Hynek, there is a significant concentration of inverted valley net- 2004]. Pedestal craters are common throughout Arabia works in the region [Fassett and Head, 2007; Williams, Terra, formed when ejecta from impact craters armored 2007] (Figure 7c). The topographic inversion of these friable sedimentary deposits, such that subsequent erosion valleys is most easily explained by the valleys having been and deflation of the surface left the crater ejecta blanket as a filled with an erosion‐resistant indurated material, and then high‐standing plateau. Pedestal craters similar to those seen left as high‐standing ridges when the surrounding terrain elsewhere in Arabia Terra are seen both beyond the margins later experienced net deflation or erosion. of the etched unit [Hynek, 2004] as well as in the process of [44] As discussed in the companion paper (manuscript in eroding out from the Meridiani plains (e.g., Figure 5a). preparation, 2010), the transition to arid conditions over [47] High resolution images from the HiRISE camera most of the surface of Mars would cause the water table to extend the morphological similarity of the Arabia Terra drop deeper beneath the surface, arresting the development deposits with the Meridiani playa deposits down to the of fluvial features. In Arabia Terra, the water table would submeter scale. Layering in some of the etched terrain de- have remained close to the surface into the Hesperian. posits exhibits a distinct periodicity (Figure 8a), similar to that During this transition, valley networks in Arabia Terra seen in Becquerel crater 1200 km to the north (Figure 8b) would see a dramatic decrease in the total flux of water, an [Lewis et al., 2008]. This periodic layering is strongly increase in the ratio of groundwater‐ to precipitation‐ suggestive of climatic control, and is consistent with a cli- sourced fluids, and a concomitant increase in the fluid matic modulation of the precipitation and evaporation flux salinity. The channels would experience a transition from driving variations in the rate groundwater upwelling and high discharge erosional flow to low discharge seepage and evaporation. Layering in other portions of the etched terrain evaporation, leading to either evaporite deposition or (Figure 8c) superficially resembles that observed in Henry cementation of aeolian debris. While inverted valley net- crater (Figure 8d), 1700 km to the southeast. Many of the works can be produced by a number of mechanisms, the intracrater deposits exhibit step‐like layering, suggestive of concentration in Arabia Terra is consistent with valley infill strength contrasts within or between layers. This step‐like by evaporites and cemented sediments due to the regional layering is observed both in outliers of the etched terrain groundwater upwelling in Arabia Terra, followed by net (Figures 8e and 8g), and in deposits within Schiapparelli deflation of the surrounding surface. crater 800 km to the east (Figure 8f). Fine‐scale layering [45] Beyond Meridiani, a widespread mantling unit is also preserved beneath the armoring crater ejecta sur- mapped throughout Arabia Terra [Greeley and Guest, 1987] rounding pedestal craters (Figure 8h). A periodicity in has been attributed to an origin as a paleo polar layered apparent layer thickness or strength similar to that expressed deposit [Schultz and Lutz, 1988; Niles and Michalski, 2009], in Becquerel (Figure 8c) is also observed in the pedestal or through aeolian, sedimentary, or volcanic processes crater in Figure 8h, and possibly in the deposits in Figures 8c [Grant and Schultz, 1990; Moore, 1990; Fassett and Head, and 8f, though these are not as clearly expressed. That a 2007]. MOC images reveal meter‐scale layering, interpreted similar periodicity in the layering is seen in deposits across as evidence for water‐lain sedimentary rocks [Malin and much of Arabia Terra further strengthens the argument for a Edgett, 2000; Edgett and Malin, 2002]. The morphology, commonality of origins. layering, and relief of the mantling units in eastern Arabia [48] As a specific comparison of the model predictions suggest deposits of dust to sand‐sized particles of aeolian or with the observed sedimentary record, we focus on two air fall origin, exhibiting variable degrees of induration or prominent degraded impact basins in eastern Arabia Terra, cementation, leading to inverted topography of craters and within which groundwater‐mediated sedimentation is pre- valleys [Fassett and Head, 2007]. A reasonable chain of dicted (see Figure 4n for context boxes). The models predict inference connects the widespread deposits across Arabia deposit thicknesses of approximately 0.5 to 1 km within Terra with those in Meridiani. Correlation of the strati- these basins. MOLA topography reveals the presence of graphic section observed by Opportunity with a widespread pedestal craters within each of these basins (Figure 9), light toned, high thermal inertia, etched unit in western preserving a remnant of eroded deposits up to 1 km thick Arabia Terra suggests that at least these deposits are that likely covered the basin floors. The apparent periodic genetically related to the playa deposits observed in Mer- bundling of layers within one of these pedestals (Figure 8h) idiani [Hynek, 2004]. OMEGA and CRISM spectra confirm suggests climatic control and a related origin to deposits the presence of sulfates in dust‐free western Arabia Terra elsewhere in Meridiani and Arabia Terra. The pedestals are [Gendrin et al., 2005; Griffes et al., 2007; Poulet et al., surrounded by smooth and ridged plains units of uncertain 2008; Wiseman et al., 2010], and similar indurated depos- origin on the basin floors. The compressional ridges on the its are seen throughout the region. The large intracrater basin floors are consistent with the plains being composed deposits across Arabia Terra are morphologically similar to of either volcanic or strongly indurated sedimentary mate- the intracrater deposit in Miyamoto crater (Figure 5a), which rials. While the original extent and thickness of the sedi- is directly continuous with the Meridiani plains. Intracrater mentary deposits throughout Arabia Terra is uncertain, these deposits in the vicinity of Meridiani Planum have been and other pedestal craters suggest that much of the region identified as outliers of the etched terrain [Hynek, 2004]. was at one time covered in sedimentary deposits in excess of 1 km in thickness.

12 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

Figure 8. HiRISE images of select representative layered deposits in Arabia Terra. The scale bar in Figure 8a is 25 m in length, and Figures 8a–8h are at the same scale. Deposits within an outlier of the Meridiani etched terrain show evidence for a periodic bundling of layers into packets (Figure 8a, image PSP_007440_1845), similar to periodic variations in layer thickness observed in Becquerel crater by Lewis et al. [2008] 1200 km to the northeast (Figure 8b, image PSP_001955_2015). Layering exposed in a crater wall within the Meridiani etched terrain (Figure 8c, image ESP_011910_1825) superficially resembles layering within an intracrater mound in Henry crater 1700 km to the east (Figure 8d, image PSP_009008_1915). Step‐like layering observed in outliers of the etched terrain (Figures 8e and 8g, images ESP_012490_1835 and PSP_002878_1880, respectively) resemble step‐like layering exposed in deposits within Schiapparelli crater 800 km to the southeast (Figure 8f (right side), PSP_002508_1800). Layering is preserved in a pedestal crater beneath an armoring layer of crater ejecta (Figure 8h, image PSP_7531_1935; see also Figure 10a). A possible periodicity in layer thickness or strength similar to that in Figures 8a and 8b is observed in Figure 8h, though the layering is partially obscured by dust and a thinner stratigraphic section is exposed. This pedestal crater is 1950 km east of the deposits shown in Figure 8a. The locations of images (Figures 8a–8h) are shown in Figure 8i, superimposed over MOLA shaded relief, with the outline of the etched terrain as identified by Hynek [2004] outlined for reference.

[49] While aeolian processes likely played a role in both since aeolian transport and deposition is ubiquitous on Mars, deposition and erosion of the deposits both within and but rather the cause of their induration and accumulation to beyond Meridiani [Fassett and Head, 2007], this does not thicknesses in excess of a kilometer in this particular region. satisfactorily explain their concentration in Arabia Terra. Our work suggests that hydrological activity leading to The key unknown is not the origin of the clastic grains, evaporite formation and groundwater‐mediated cementation

13 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

of aeolian debris played a pivotal role in the origin of the Arabia Terra deposits. As aeolian activity continuously redistributed sand and dust across the surface of the planet, these materials would have been infiltrated by groundwater in Arabia Terra. Evaporation of the groundwater at the sediment surface would have precipitated sulfate salts, cementing the materials in place. The resulting gradual aggradation of the surface would have been accompanied by a parallel rise in the water table, allowing accumulation of thick deposits (Figure 10). As the hydrological cycle waned in the early Hesperian, precipitation and evaporation rates would have decreased and the water table would have dropped deeper beneath the surface. The sedimentary deposits across Arabia Terra would then be left above the water table, preventing further induration of the wind‐ transported sand and dust, and allowing the net deflation of the region to begin. Deposits were preferentially preserved where protected from subsequent erosion by either a lag deposit of hematite concretions or by an armoring layer of crater ejecta.

6. Alternative Interpretations of the Meridiani Planum Deposits

[50] This study focuses on understanding the mechanism required to maintain the groundwater upwelling and playa environment inferred from the rover observations, in con- trast with other studies that propose different interpretations of the observations and alternative depositional environments and mechanisms [Knauth et al., 2005; McCollom and Hynek, 2005; Niles and Michalski, 2009]. Debate continues over the interpretation of the Meridiani deposits, their environment of deposition, and the ultimate source of the sulfates. Within the context of the playa interpretation, we have shown that regional to global scale groundwater flow can explain the location, thickness, dip angle and azimuth, and deposition rate of the deposits, while also making testable predictions regarding the distribution of deposits throughout the Arabia Terra region. [51] Proposed depositional mechanisms involving volcanic fumaroles [McCollom and Hynek, 2005] or impact base surges [Knauth et al., 2005] offer alternative mechanisms for producing the deposit morphology and mineralogy, but do not address the spatial distribution. Impacts are ubiquitous on Mars, and there is no evidence to suggest that impact base surges should take on a different form in Meridiani and Arabia Terra than elsewhere. Furthermore, an impact origin is difficult to reconcile with the observation that the deposits postdate the largest craters in the region and that many craters are filled to their rims with deposits. Figure 9. (a–b) Two large degraded impact basins in east- Similarly, while volcanic processes have been widespread ern Arabia Terra, for which the model predicts significant on Mars over its history, the observed deposits do not have accumulations of evaporites and cemented sediments (insets, any clear relationship to known volcanic vents. For both see Figure 4n for color scale and context boxes). The basins impact and volcanic interpretations, some special circum- are floored with smooth to ridged plains materials of uncer- stance must be invoked in order to explain the localization tain origin. (c–d) Pedestal craters within the basins preserve of the deposits to Arabia Terra. Furthermore, these mechan- the remnants of formerly more extensive deposits, suggesting isms do not adequately explain the association of similar preerosional deposit thicknesses within the basins of up to sulfate and hematite deposits with regions of unambiguous 1 km. The location of Figure 8h, which reveals periodic groundwater outflow in the chaos regions and Valles Mar- bundles of layers, is shown in Figure 9a. ineris canyons [Gendrin et al., 2005; Glotch and Christensen, 2005; Glotch and Rogers, 2007; Noe Dobrea et al., 2008;

14 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

Michalski [2009] argued that closed system weathering provides the best explanation for the elemental abundances. They note that the scatter in elemental abundance data is reasonably fit by simply adding or subtracting MgSO4. The fact that the majority of the chemical variation within the Meridiani outcrops is best explained by MgSO4 mobility is expected, since epsomite (MgSO4 ·7H2O) is the last and most soluble mineral precipitated during evaporation of basaltic weathering fluids in thermodynamic modeling [Tosca et al., 2005]. The composition of the deposits would have been largely homogenized by mixing of grains during aeolian transport and redeposition within the paleo dune field. Later episodes of groundwater‐mediated diagenesis would preferentially dissolve MgSO4 from the sediments lower in the sequence, resulting in the observed decrease in Mg and S with depth [Squyres et al., 2009]. [53] The key question is the nature and origin of the material prior to this late diagenesis. Niles and Michalski [2009] note that the relative cation abundances aside from Mg are close to those in unaltered basalt, which they interpret as evidence for closed system alteration within an ice cap. A relatively homogenous composition is consistent with the broad regions of nearly uniform groundwater upwelling predicted by the hydrological models (Figure 4o). The Opportunity rover has sampled only a small fraction of the full horizontal and vertical extent of the Meridiani deposit of approximately 1000 km and 1 km, respectively. Closed system alteration of basaltic grains would require that the entire Meridiani deposit have basaltic cation abundances. While observations are limited by the range of the Opportunity rover and the vertical exposure of the outcrops, elemental abundance data from Victoria Crater reveals Fe enrichments relative to other elements of approximately 20% in comparison with the outcrops at Endurance Crater Figure 10. Cartoon diagram of sequence of events in the [Squyres et al., 2009]. Orbital observations reveal lateral evolution of the Arabia Terra deposits. The water table is de- mineralogical heterogeneity within Meridiani, with varia- noted by the blue line and marked with an inverted triangle. tions in the hydration state of sulfates, strength of the sulfate Ancient crust is shown in gray, and sedimentary deposits signature in the spectra, and degree of induration [Wiseman shown in beige with layering in black. (a) Early groundwa- et al., 2010]. These orbital observations are consistent with, ter evaporation is limited to the largest impact craters. (b–c) but do not require, significant heterogeneity in the elemental As the craters fill with sediments, the water table rises until composition of the deposits as well. the groundwater‐driven playa expands laterally over the [54] Perhaps the strongest argument for an open system intercrater plains. (d) Portions of the deposit become origin for the deposits involving groundwater is the identi- armored by ejecta from impacts or by the accumulation of fication of spectrally and geomorphically similar sulfate‐ a lag of hematite concretions as in Meridiani, or may simply hematite deposits at the sources of the outflow channels in be more resistant to erosion due to a greater degree of indu- chaos regions and the Valles Marineris canyons [Gendrin ration. (e) A subsequent drop in the water table exposes the et al., 2005; Glotch and Christensen, 2005; Glotch and deposits to aeolian deflation, leaving behind thick intracrater Rogers, 2007; Noe Dobrea et al., 2008; Weitz et al., 2008; deposits and isolated remnants of deposits in the form of Lichtenberg et al., 2010]. A closed system source for these pedestal craters or other plains‐covering deposits. deposits is at odds with the unambiguous geomorphic evidence for the release of enormous quantities of groundwater Weitz et al., 2008; Lichtenberg et al., 2010], which argue from these regions. The vast amounts of groundwater strongly for a groundwater origin for the Meridiani deposits. required to carve the outflow channels requires that the chaos [52] Several arguments against the playa hypothesis for regions tap aquifers of substantial regional extent [Andrews‐ the origin of the deposits have been framed around the Hanna and Phillips, 2007]. elemental abundance data from the rover observations. [55] It has also been argued that deeply sourced fluids in a McCollom and Hynek [2005] argued that the elemental basaltic aquifer are inconsistent with the inferred acidity of abundances observed in the Meridiani outcrops were best the fluids at the surface [Niles and Michalski, 2009]. While matched by the addition of pure SO2 to unweathered basalt, the fluids may indeed be buffered at a near‐neutral pH in the though Squyres et al. [2006] demonstrated that the data was deep subsurface, as the slowly upwelling groundwater nears best fit by a mixture of weathered basalt and (Mg, Fe, Ca) the oxidized surface environment, oxidation of the ground- SO4 as predicted by the playa hypothesis. Niles and water would be accompanied by increasing acidity [Burns,

15 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

1993; Burns and Fisher, 1993; Hurowitz et al., 2010]. western Hellas and other regions at the time of formation of Within the playa, the rise and fall of the water table and the Hesperian outflow channels [Harrison and Grimm, contemporaneous aeolian reworking of the sediments would 2009]. While the assumption of a globally interconnected be highly effective in oxidizing both the sediments and aquifer system is not appropriate for the Earth, such large‐ fluids near the surface. Evaporitic concentration of the scale interconnection on Mars is suggested by the uniform basaltic weathering fluids would continue to lower the pH impact bombardment of the southern highlands, which [Tosca et al., 2005]. Furthermore, the presence of Fe/Mg would have intensely fractured the crust over the entire smectites lower in the stratigraphic column argues for for- surface of Noachian Mars [Clifford, 1981, 1993]. We see no mation of some portions of the deposit under higher pH evidence for impermeable aquicludes subdividing aquifers conditions [Wiseman et al., 2010]. within the southern highlands. The lack of evidence for [56] Previous studies have raised arguments against the outflows within Hellas may have been a result of a possible playa hypothesis on the basis of the large horizontal extent ice‐covered sea within the basin at the time [Moore and of the deposits, the large volumes of groundwater required, Wilhelms, 2001], or may suggest the importance of more and the lack of an enclosed basin [McCollom and Hynek, local pressurization mechanisms at the sources of the out- 2005; Niles and Michalski, 2009]. This work addresses flow channels during the Hesperian. However, a source for each of these objections. We have shown that regional‐scale the Meridiani fluids from regional groundwater flow does groundwater flow patterns would have maintained a playa not require a globally interconnected aquifer, rather only environment not only in Meridiani Planum, but also across that aquifers be interconnected from the hydraulic divide much of Arabia Terra. Growth of the deposits smoothes between Arabia Terra and Hellas through to the proximal the surface topography, allowing horizontally continuous portions of the northern lowlands. Indeed the strong agree- deposits over distances of many 100s of km, with consistent ment between predicted and observed loci of groundwater dip directions approximately following the regional slopes, upwelling in Arabia Terra provides compelling evidence in as found by Hynek and Phillips [2008]. The regional nature favor of long distance groundwater flow and aquifers that of the groundwater flow would also allow for consistent were interconnected on at least regional scales on early composition of the deposits across great distances. The Mars. groundwater upwelling in the absence of an enclosed basin is driven by the unique topography of Arabia Terra relative to the highlands and lowlands. The predicted rates of 7. Discussion groundwater upwelling could generate the observed deposit ‐ thickness in timescales of 10s to 100s of Myr, depending on [59] These results show that late Noachian early Hespe- the assumed fluid salinity and aquifer properties. rian evaporitic sulfate deposits in Meridiani Planum were a ‐ [57] It has been suggested that the groundwater discharge natural result of groundwater upwelling in a precipitation ‐ at Meridiani Planum may have resulted from Tharsis‐ evaporation driven hydrological cycle operating under arid induced tilting of the crust and lithosphere in Arabia Terra conditions. Meridiani and the surrounding Arabia Terra [Hynek and Phillips, 2005; Arvidson et al., 2006b]. How- region are among the few regions of presently exposed ‐ ever, in previous work we investigated the hydrological Noachian aged crust for which sustained groundwater evolution of Mars before, during, and after Tharsis forma- upwelling and evaporation is predicted. A steady flux of tion, and found Meridiani Planum to emerge as a region of groundwater to the surface would have sustained regional focused groundwater upwelling in all three cases [Andrews‐ playa development in the absence of a topographic basin, Hanna et al., 2007]. The groundwater upwelling is driven driven by the combination of the general low topography of rather by a combination of a precipitation‐evaporation‐ Arabia Terra and the distinct breaks in slope separating it driven hydrological cycle and the regional topography of from the southern highlands and northern lowlands. Arabia Terra. The slopes in Arabia Terra are primarily a Groundwater evaporation would have produced deposits of result of isostatically compensated crustal thickness varia- substantial thickness, both through the direct precipitation of tions [e.g., Neumann et al., 2004]. Our loading models evaporites and the cementation and induration of aeolian suggest that the modeled Tharsis‐induced changes in slope sediments. Results suggest a temporal evolution, in which outside the rise are one to two orders of magnitude less than groundwater upwelling is first restricted to isolated depres- the preexisting slopes associated with the isostatic crustal sions and craters, but then spreads laterally across the sur- thickness variations of the dichotomy and Arabia Terra face as these depressions fill with deposits. [Andrews‐Hanna et al., 2007]. Localized admittance values [60] Deposits similar to the Meridiani sediments are pre- in Arabia Terra require a thin lithosphere at the time of dicted to be widespread across much of Arabia Terra. formation [Evans et al., 2010], suggesting that the low Multiple lines of evidence suggest that Arabia Terra was a topography of this region was established early in Mars nexus of hydrological activity throughout the Noachian and ‐ history. This is consistent with the interpretation of Arabia early Hesperian epochs. Observations of sediment filled Terra as a partial ring structure around the Borealis impact craters, inverted valley networks, widespread layered man- basin [Andrews‐Hanna et al., 2008]. While Tharsis tling deposits, pedestal craters, and the Meridiani playa undoubtedly did affect the hydrological evolution of Mars, it evaporites are all consistent with the model prediction of did not play a dominant role in maintaining the Meridiani sustained groundwater upwelling and evaporation in Arabia playa. Terra. Variations in the morphology and composition of the [58] It has been suggested that groundwater flow on Mars deposits likely result from varying proportions of evaporitic was only regional in extent, based on the lack of outflow and clastic materials, possibly driven by variations in either features in regions of predicted groundwater outflow in the groundwater flux or sediment supply. The hydrological

16 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002 models successfully explain the spatial distribution, rate of tinuous flow and a smooth variation in hydraulic head deposition, thickness, and dip of the Arabia Terra deposits. across the poles. [61] This long‐wavelength groundwater flow is similar to [64] The aquifer porosity and hydraulic conductivity were that found in arid regions of the Earth, such as the Great taken from the megaregolith model of Hanna and Phillips Artesian Basin of Australia [Habermehl, 1980]. This [2005]. The depth dependence of the aquifer properties in hydrological cycle requires active precipitation‐induced this model is generally similar to other models [Clifford, recharge of the low latitude aquifers, though this precipita- 1993; Clifford and Parker, 2001; Wang et al., 2006], and tion could have been in the form of either rain or snow, our investigation of a range of parameter space shows that provided that the snow was able to melt and infiltrate the the results of this study are not strongly sensitive to the surface. The regional to global nature of the hydrological specific aquifer model. The discontinuity in the porosity at cycle is not consistent with spatially and temporally isolated the base of the megaregolith at 2 km depth in the aquifer impact‐induced greenhouse climates [Segura et al., 2002], model of Hanna and Phillips [2005] is smoothed to a cosine or localized volcanic warming of the subsurface [McCollom transition between 1 and 3 km in order to improve the model and Hynek, 2005]. While local hydrothermal circulation can stability. The aquifer properties are updated to reflect the elevate the hydraulic head over small regions, the hydro- changing hydraulic head once every 10 time steps, which logical cycle responsible for Meridiani evaporite formation decreases the model run time without adversely affecting requires long distance cycling of water between the low- performance. lands and highlands, and is incompatible with the typical [65] Global models were typically run at a spatial reso- length scales of hydrothermal groundwater circulation lution of 2.5 degrees per pixel, a limitation imposed by the [Travis et al., 2003]. At the locations of groundwater model run times. These models are at a higher resolution upwelling, the formation of playa evaporites requires tem- than the 5 degrees per pixel results presented earlier peratures above the freezing point of the groundwater [Andrews‐Hanna et al., 2007]. Higher resolution simula- brines. Conditions conducive to both precipitation‐induced tions were performed globally at 1.25 degrees per pixel, and aquifer recharge and groundwater evaporation must have locally at 0.25 degrees per pixel over a limited region, as been stable or episodically present over long periods of time discussed in the main text. (likely 10s to 100s of Myr) over much of the surface of Mars in order to establish and maintain the hydrological cycle. A groundwater‐driven playa origin for the Meridiani deposits thus implies that conditions conducive to life persisted Appendix B: Model Benchmarking across much of the low latitudes to midlatitudes of Mars as [66] The full model representing two‐dimensional late as the early Hesperian. unconfined groundwater flow in a spherical shell with [62] The detailed in situ observations by the Mars depth‐dependent hydraulic conductivity and porosity is not Exploration Rover Opportunity reveal the specific local amenable to analytic solutions. Benchmarking was per- depositional environment and modification history of the formed with a simplified model in an aquifer with uniform Meridiani deposits: that they are composed of dirty eva- porosity and hydraulic conductivity. In the limit of high porites that have been substantially reworked by aeolian and model resolution over regions of limited size, the effect of fluvial processes and diagenetically modified by a fluctu- the planetary curvature on the flow will be unimportant and ating water table, consistent with formation in a playa. The the model results should be reasonably approximated by hydrological modeling results presented here reveal the analytic solutions in one‐ or two‐dimensional Cartesian deeper mechanisms at work and the broader hydrologic and geometries. We applied constant head boundary conditions climatic implications of the deposits: that they formed as a at two ends of a one‐dimensional strip of constant elevation result of sustained regional groundwater upwelling in within the model, subject to lateral no‐flow boundary con- response to a precipitation‐evaporation‐driven hydrological ditions and a constant rate of recharge at the surface. For cycle operating in an arid climate, with the location of the unconfined flow under Dupuit assumptions, the hydraulic deposits resulting from the unique topography of the sur- head should follow a parabolic profile, and the model results rounding Arabia Terra region. This work highlights the agreed with the analytic solutions to the percent level. The importance of combining in situ and orbital observations model was also benchmarked against two‐dimensional with a theoretical understanding of the hydrological cycle. analytic solutions for unconfined aquifer drawdown surrounding a well in which a constant pumping rate is applied [Demming, 2002], with the errors again less than the percent Appendix A: Model Details level. The model was benchmarked against simpler 1‐D radially symmetric numerical models, showing perfect [63] The hydrological model represents unconfined agreement. As a test of the spherical geometry of the model, groundwater flow using the Dupuit approximation the diffusion of a hemispheric step function in hydraulic [Demming, 2002], which assumes that the hydraulic gra- head was repeated with the step function oriented either dients are sufficiently small that flow lines are horizontal north‐south or east‐west, demonstrating that the longitudi- and vertical flow can be neglected. These assumptions are nal flow across the model edges and latitudinal flow across ‐ reasonable approximations for the long wavelength the poles are accurately represented in the models. groundwater flow of interest here. Equation 1 was dis- [67] While the full model cannot be benchmarked against ‐ cretized over a uniform latitude longitude grid. Additional an analytic solution, it can be tested for conservation of “ ” cap cells were added at the poles in hydrologic commu- water volume. In equilibrium, continued flow driven by the nication with all cells in the adjacent row, allowing con- precipitation and evaporation boundary conditions should

17 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002 maintain a constant volume of water in the aquifers. The [70] We also consider a model in which the vertical per- total water inventory was tracked during model simulations, meability structure within the aquifer is changed, with the and a small numerical drift was found to result, likely from a ratio of the permeability at the surface to that at a depth of combination of the discretization of the equations of flow 10 km decreased by a factor of 10. This is implemented in and numerical rounding errors. This drift was found to be terms of the hydrological model of Hanna and Phillips first order in time and second order in space, as would be [2005] by decreasing the maximum aperture of fractures at expected for the fully explicit finite difference formulation. the surface while holding the minimum aperture at depth For short‐term simulations (e.g., the one‐time diffusion of a constant. This change in the vertical permeability structure hydraulic step function), this numerical drift was inconse- has a noticeable effect on the global distribution of quential to the results. However, in the long term simula- groundwater and surface evaporation. In particular, a greater tions with sustained evaporation and recharge, the numerical drawdown of the water table is observed around the Hellas drift would accumulate to nontrivial levels on timescales of impact basin than observed in the baseline model. The more ∼100 Myr, effectively changing the total water inventory rapid decrease in permeability with depth in the baseline over time. In order to allow for long‐term simulations, the model impedes the aquifer drawdown around Hellas. As the total fluid volume within the aquifers was tracked during the water table drops deeper beneath the surface, the mean simulations and a correction term was added to the precip- aquifer permeability decreases, thereby impeding flow itation in order to maintain conservation of fluid volume. toward the basin and increasing the concavity of the draw- This drift correction generally modified the precipitation rate down cone. However, the predicted distribution of sedi- by less than 1%. Increasing the spatial or temporal resolu- mentary deposits in Arabia Terra remains largely the same tion decreased the drift correction, but also resulted in in both models. increased model run times. Since this correction term is [71] Model results would also likely be affected by hori- much less than expected natural spatial and temporal var- zontal anisotropy in the aquifer permeability. Orbital ob- iations in the distribution and rate of precipitation, we expect servations reveal evidence for water‐mediated bleaching it has a negligible effect on the results and conclusions. around deformation bands in Arabia Terra and elsewhere [Okubo and McEwen, 2007; Okubo et al., 2007]. Com- pressional deformation bands may act as barriers to the flow, while tensile or shear deformation bands may act as conduits Appendix C: Parameter Sensitivity to the flow, with both tending to channel groundwater par- [68] Martian hydrological models are inherently under allel to the structures. The horizontal stress directions in the constrained. We have no direct data relevant to the hydro- Martian lithosphere are dominated by Tharsis loading, with logical properties of the subsurface, or how those properties the greatest (compressional) principle stress direction gen- vary in space and time. We employ the hydrological model erally oriented radial to Tharsis outside of the rise, leading to of Hanna and Phillips [2005], which estimates likely ranges circumferential wrinkle ridges. Resulting tectonism and of the porosity, permeability, and compressibility using a deformation of the crust would act to deflect groundwater ‐ combination of terrestrial measurements, and terrestrial and flow in a north south direction in Arabia Terra. As a result, lunar analog materials. In that model, the permeability groundwater may follow paths oblique to the regional ranges from 10−15 to 10−11 m2, the porosity ranges from hydraulic gradients. However, because of the large hori- 0.16 to 0.04, and the compressibility ranges from 2 × 10−9 to zontal extent of Arabia Terra and the consistency of slopes 1.5 × 10−10 Pa−1 between the surface and a depth of 10 km. across this region, we expect the general conclusions of this These values are broadly similar to those in other hydro- study to be unchanged for the case of aquifer anisotropy. logical models of Mars [Clifford, 1993; Clifford and Parker, [72] Lateral variability in the aquifer properties may have 2001; Wang et al., 2006] and the Earth [Manning and also influenced the groundwater flow patterns and resulting Ingebritsen, 1999; Saar and Manga, 2004]. Of these para- deposits. Adopting an approach similar to that of Harrison meters, the greatest uncertainty lies in the permeability, and Grimm [2009], we have investigated the effect of spa- which varies by several orders of magnitude within terres- tially varying aquifer properties. As in that study, the aquifer trial aquifers. The permeability determines the equilibrium hydraulic conductivity in each model cell was scaled by a hydrological state of the planet, while the aquifer porosity spatially varying factor from the nominal value for that cell. and compressibility (for the case of confined aquifers) affect Following Harrison and Grimm [2009], we construct this only the time‐dependent hydrological response. scaling function as a randomly varying logarithmic factor [69] The patterns of groundwater flow and distribution of with a variance of 0.5 and a correlation length of 15° on the groundwater upwelling is insensitive to a uniform scaling of surface (Figure C1a). A total of 30 realizations of this spa- any of the relevant aquifer properties. For example, if the tially varying random scaling function were generated, and permeability is increased globally by a constant factor, this the hydraulic evolution simulated for each. The model run scaling factor will come out in front of the derivatives in the times were accelerated by a factor of 5 by artificially first term in equation (1) and the precipitation rate increases increasing the ratio of the deposition rate to the evaporation by a similar factor, resulting only in a scaling of the rate of rate, which was found to have a minimal effect on the re- temporal evolution of the system. This was confirmed in a sulting evaporite distribution. The hydraulic conductivity model with a tenfold increase in permeability, with the results scaling factor and resulting deposit distribution from a sin- after 5 Myr of evaporite deposition being identical within gle randomly selected model are shown in Figures C1a and the numerical precision to those in the nominal model after C1b. The mean (±one standard deviation) deposit thickness 50 Myr. agrees well with that predicted by the uniform aquifer model (Figures C1c–C1e). The mean deposits are more broadly

18 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

Figure C1. Effect of spatially varying aquifer hydraulic conductivity. (a) The log10 of the hydraulic conductivity scaling factor for a single realization of the laterally heterogeneous aquifer model. (b) The predicted deposit thickness from this single model and (c–e) the mean and ± 1 standard deviation deposit thickness for the full suite of 30 models are consistent with the predictions for the globally uniform model (Figure 2k). (f) The model outcome that deviates most strongly from the mean distribution. distributed across the surface due to the variations in loci of predict significant evaporite deposition in Meridiani Planum groundwater upwelling from one model to another. The and the Arabia Terra region. Modest variations from one results of the single model outcome that differs most from model to another provide both better and worse matches to the mean distribution of deposits is presented in Figure C1f. the observed deposit distribution. In general we find that Interestingly, this most deviant model results in a thicker lateral heterogeneity in the aquifer properties results in only deposit more localized to the Meridiani etched terrain than modest changes to the predicted distribution of deposits. predicted by the uniform aquifer model. All models clearly Given the lack of constraints on the actual spatial vari-

Figure C2. (a–c) Predicted depth to the water table, evaporite thickness, and evaporation rate for different assumed precipitation distributions, including a cosine distribution between ± 45° latitude, (d–f) a uniform distribution between ± 45° latitude, (g–i) a cosine distribution between ± 60° latitude, (j–l) and a cosine distribution between ± 45° latitude scaled by a factor proportional to the local elevation. Similar distributions of evaporite deposits are predicted in all cases.

19 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002 ability, the results of the uniform aquifer model are most Arvidson, R. E., et al. (2006a), Nature and origin of the hematite‐bearing representative of the likely distribution of deposits. plains of Terra Meridiani based on analyses of orbital and Mars Explo- ration Rover data sets, J. Geophys. Res., 111, E12S08, doi:10.1029/ [73] The assumed distribution of precipitation on the 2006JE002728. surface can also affect the hydrological evolution. The Arvidson, R. E., S. W. Squyres, B. C. Clark, J. P. Grotzinger, A. H. Knoll, nominal model distributed the precipitation in a longitudi- S. M. McLennan, and N. J. Tosca (2006b), Regional setting and model nally uniform belt following a half cosine distribution with for the Meridiani planum deposits investigated by the Opportunity Mars Exploration Rover, Lunar Planet. Sci., XXXVII, Abstract 1400. latitude between ± 45°. We have performed simulations Barnhart, C. J., A. D. Howard, and J. M. Moore (2009), Long‐term precip- with a uniform distribution between ± 45°, as well as a half itation and late‐stage valley network formation: Landform simulations of cosine distribution between ± 60° latitude. Finally, we the Parana Basin, Mars, J. Geophys. Res., 114, E01003, doi:10.1029/ 2008JE003122. considered a model in which the precipitation rate depends Bibring, J.‐P., et al. (2006), Global mineralogical and aqueous Mars history on both latitude and elevation, with the half cosine precip- derived from OMEGA/Mars Express data, Science, 312,400–404, itation distribution between ± 45° latitude scaled by a factor doi:10.1126/science.1122659. Boynton, W. V., et al. (2002), Distribution of hydrogen in the near surface varying linearly between elevations of 2 and 5 km. Some of Mars: Evidence for subsurface ice deposits, Science, 297,81–85, elevation dependence of the precipitation rate is suggested doi:10.1126/science.1073722. by the observation that valley network densities are greater Brain, D. A., and B. M. Jakosky (1998), Atmospheric loss since the onset at higher elevations [Carr, 1995]. However, this relationship of the Martian geologic record: Combined role of impact erosion and sputtering, J. Geophys. Res., 103,22,689–22,694, doi:10.1029/ is likely biased by the apparent lack of valley networks in 98JE02074. the low elevation Arabia Terra region. Subsequent studies Burns, R. G. (1993), Rates and mechanisms of chemical weathering of found evidence for inverted channels in this region re- ferromagnesium silicate minerals on Mars, Geochim. Cosmochim. Acta, 57, 4555–4574, doi:10.1016/0016-7037(93)90182-V. presenting a population of buried and exhumed valley net- Burns, R. G., and D. S. Fisher (1993), Rates of oxidative weathering on the works [Williams, 2007]. In all models, we observe some surface of Mars, J. Geophys. Res., 98,3365–3372, doi:10.1029/ variability in the predicted depth to the water table, but the 92JE02055. global patterns remain the same (Figure C2). Only minor Carr, M. H. (1995), The Martian drainage system and the origin of valley networks and fretted channels, J. Geophys. Res., 100, 7479–7507, changes in the predicted distribution of deposits are doi:10.1029/95JE00260. observed, with all models predicting significant evaporite Carr, M. H., and F. C. Chuang (1997), Martian drainage densities, deposits in Arabia Terra. J. Geophys. Res., 102, 9145–9152, doi:10.1029/97JE00113. Carr, M. H., and J. W. Head (2003), Basal melting of snow on early Mars: [74] Models with resolutions ranging from 2.5 degrees per A possible origin of some valley networks, Geophys. Res. Lett., 30(24), pixel to 0.25 degrees per pixel predict similar distributions 2245, doi:10.1029/2003GL018575. of groundwater, but different precipitation rates as discussed Carr, M. H., and M. C. Malin (2000), Meter‐scale characteristics of Martian channels and valleys, Icarus, 146, 366–386, doi:10.1006/icar.2000.6428. in the text. Differences in the regional water table heights of Christensen, P. R., et al. (2000), Detection of crystalline hematite mineral- a few hundred meters are observed prior to any evaporite ization on Mars by the Thermal Emission Spectrometer: Evidence for deposition, driven by the effects of smaller craters in the near surface water, J. Geophys. Res., 105, 9623–9642, doi:10.1029/ high resolution models as discussed above. As evaporite 1999JE001093. Christensen, P. R., R. V. Morris, M. D. Lane, J. L. Bandfield, and deposition smoothes the short wavelength topography in M. C. Malin (2001), Global mapping of Martian hematite mineral deposits: regions of active evaporite deposition, the higher resolution Remnants of water‐driven processes on early Mars, J. Geophys. Res., 106, model results converge with the lower resolution models. 23,873–23,885, doi:10.1029/2000JE001415. Clifford, S. M. (1981), A pore volume estimate of the Martian megaregolith [75] This parameter investigation demonstrates that the based on a lunar analog, LPI Contrib., 441,46–48. general conclusions of this study are robust despite large Clifford, S. M. (1993), A model for the hydrologic and climatic behavior of uncertainties in the aquifer properties and the ancient pre- water on Mars, J. Geophys. Res., 98, 10,973–11,016, doi:10.1029/ cipitation distribution. The one parameter that does exert a 93JE00225. Clifford,S.M.,andT.J.Parker(2001),TheevolutionoftheMartian dominant control on the hydrological evolution of Mars is hydrosphere: Implications for the fate of a primordial ocean and the current the total water inventory (expressed in terms of the initial state of the northern plains, Icarus, 154,40–79, doi:10.1006/ mean water table depth in the models). However, rather than icar.2001.6671. Cloutis, E. A., et al. (2006), Detection and discrimination of sulfate minerals being a simple source of uncertainty, this a parameter likely using reflectance spectroscopy, Icarus, 184,121–157, doi:10.1016/j.icarus. experienced real secular and periodic variations with time 2006.04.003. due to climatic forcing, as discussed in the companion paper Craddock, R. A., and A. D. Howard (2002), The case for rainfall on a warm, wet early Mars, J. Geophys. Res., 107(E11), 5111, doi:10.1029/ (manuscript in preparation, 2010). 2001JE001505. Demming, D. (2002), Introduction to Hydrology, 468 pp., McGraw Hill, [76] Acknowledgments. This work was supported by a grant to J.C. ‐ New York. A H by the NASA Mars Data Analysis Program. We are grateful to Joyce Edgett, K. S., and M. Malin (2002), Martian sedimentary rock stratigraphy: Hoopes for assistance in compiling the HiRISE images used in this paper. Outcrops and interbedded craters of northwest Sinus Merdiani and south- This manuscript benefited from thorough and thoughtful reviews by Tim west Arabia Terra, Geophys. Res. Lett., 29(24), 2179, doi:10.1029/ Glotch and an anonymous reviewer. 2002GL016515. Evans, A. J., J. C. Andrews‐Hanna, and M. T. Zuber (2010), Geophysical limitations on the erosion history within Arabia Terra, J. Geophys. Res., doi:10.1029/2009JE003469, in press. References Fassett, C. I., and J. W. Head (2007), Layered mantling deposits in north- Andrews‐Hanna, J. C., and R. J. Phillips (2007), Hydrological modeling of east Arabia Terra, Mars: Noachian‐Hesperian sedimentation, erosion, outflow channels and chaos regions, J. Geophys. Res., 112, E08001, and terrain inversion, J. Geophys. Res., 112, E08002, doi:10.1029/ doi:10.1029/2006JE002881. 2006JE002875. Andrews‐Hanna, J. C., R. J. Phillips, and M. T. Zuber (2007), Meridiani Fassett, C. I., and J. W. Head (2008), Valley network‐fed, open‐basin lakes Planum and the global hydrology of Mars, Nature, 446,163–166, on Mars: Distribution and implications for Noachian surface and subsur- doi:10.1038/nature05594. face hydrology, Icarus, 198,37–56, doi:10.1016/j.icarus.2008.06.016. Andrews‐Hanna, J. C., M. T. Zuber, and W. B. Banerdt (2008), The Feldman, W. C., et al. (2004), Global distribution of near‐surface hydrogen Borealis Basin and the origin of the Martian crustal dichotomy, on Mars, J. Geophys. Res., 109, E09006, doi:10.1029/2003JE002160. Nature, 453, 1212–1215, doi:10.1038/nature07011.

20 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

Gendrin, A., et al. (2005), Sulfates in Martian layered terrains: The Manning, C. E., and S. E. Ingebritsen (1999), Permeability of the continen- OMEGA/Mars Express view, Science, 307, 1587–1591, doi:10.1126/ tal crust: Implications of geothermal data and metamorphic systems, Rev. science.1109087. Geophys., 37(1), 127–150, doi:10.1029/1998RG900002. Glotch, T. D., and J. L. Bandfield (2006), Determination and interpretation McCollom, T. M., and B. M. Hynek (2005), A volcanic environment for of surface and atmospheric Miniature Thermal Emission Spectrometer bedrock diagenesis at Meridiani Planum on Mars, Nature, 438,1129– spectral end‐members at the Meridiani Planum landing site, J. Geophys. 1131, doi:10.1038/nature04390. Res., 111, E12S06, doi:10.1029/2005JE002671. McLennan, S. M., et al. (2005), Provenance and diagenesis of the evaporite‐ Glotch, T. D., and P. R. Christensen (2005), Geologic and mineralogic bearing Burns formation, Meridiani Planum, Mars, Earth Planet. Sci. Lett., mapping of Aram haos: Evidence for a water‐rich history, J. Geophys. 240,95–121, doi:10.1016/j.epsl.2005.09.041. Res., 110, E09006, doi:10.1029/2004JE002389. Moller, P., et al. (1997), Paleofluids and recent fluids in the upper continen- Glotch, T. D., and A. D. Rogers (2007), Evidence for aqueous deposition of tal crust: Results from the German Continental Deep Drilling Program hematite‐ and sufate‐rich light‐toned layered deposits in Aureum and (KTB), J. Geophys. Res., 102, 18,233–18,254, doi:10.1029/96JB02899. Iani chaos, Mars, J. Geophys. Res., 112, E06001, doi:10.1029/ Moore, J. M. (1990), Nature of the mantling deposit in the heavily cratered 2006JE002863. terrain of northeastern Arabia, Mars, J. Geophys. Res., 95, 14,279– Golombek, M. P., J. A. Grant, L. S. Crumpler, R. E. Arvidson, J. F. Bell, 14,289. C. M. Weitz, R. J. Sullivan, P. R. Christensen, L. A. Soderblom, and Moore, J. M., and D. E. Wilhelms (2001), Hellas as a possible site of S. W. Squyres (2006), Erosion rates at the Mars Exploration Rover landing ancient ice‐covered lakes on Mars, Icarus, 154,258–276, doi:10.1006/ sites and long‐term climate change on Mars, J. Geophys. Res., 111, icar.2001.6736. E12S10, doi:10.1029/2006JE002754. Murchie, S. L., et al. (2009), Evidence for the origin of layered deposits Grant, J. A., and P. H. Schultz (1990), Gradational epochs on Mars: in Candor Chasma, Mars, from mineral composition and hydrologic Evidence from west‐northwest of Isidis basin and Electris, Icarus, modeling, J. Geophys. Res., 114, E00D05, doi:10.1029/2009JE003343. 84, 166–195, doi:10.1016/0019-1035(90)90164-5. Neumann, G. A., M. T. Zuber, M. A. Wieczorek, P. J. McGovern, Greeley, R., and J. E. Guest (1987), Geologic map of the eastern equatorial F. G. Lemoine, and D. E. Smith (2004), Crustal structure of Mars from region of Mars, U.S. Geol. Surv. Misc. Invest. Map,I‐1802‐B. gravity and topography, J. Geophys. Res., 109, E08002, doi:10.1029/ Griffes, J. L., R. E. Arvidson, F. Poulet, and A. Gendrin (2007), Geologic 2004JE002262. and spectral mapping of etched terrain deposits in northern Meridiani Niles, P. B., and J. Michalski (2009), Meridiani Planum sediments on Planum, J. Geophys. Res., 112, E08S09, doi:10.1029/2006JE002811. Mars formed through weathering in massive ice deposits, Nat. Geosci., Grotzinger, J. P., et al. (2005), Stratigraphy and sedimentology of a dry to 2, 215–220, doi:10.1038/ngeo438. wet eolian depositional system, Burns formation, Meridiani Planum, Noe Dobrea, E. Z., F. Poulet, and M. C. Malin (2008), Correlations Mars, Earth Planet. Sci. Lett., 240,11–72, doi:10.1016/j.epsl.2005. between hematite and sulfates in the chaotic terrain east of Valles Mari- 09.039. neris, Icarus, 193, 516–534, doi:10.1016/j.icarus.2007.06.029. Habermehl, M. A. (1980), The Great Artesian Basin, Australia, J. Aust. Okubo, C. H., and A. S. McEwen (2007), Fracture‐controlled paleo‐fluid flow Geol. Geophys., 5,9–38. in Candor Chasma, Mars, Science, 315, 983–985, doi:10.1126/science. Handford, C. R. (1991), Marginal marine halite: Sabkhas and salinas, in 1136855. Evaporites, Petroleum, and Mineral Resources, edited by J. L. Melvin, Okubo, C. H., et al. (2007), Deformation bands on Mars and implications pp. 1–66, Elsevier, New York. for subsurface fluid flow, in Seventh International Conference on Mars, Hanna, J. C., and R. J. Phillips (2005), Hydrological modeling of the July 9–13, 2007, Pasadena CA [CD‐ROM], LPI Contrib., 1353, abstract Martian crust with application to the pressurization of aquifers, J. Geophys. 3041. Res., 110, E01004, doi:10.1029/2004JE002330. Pelkey, S. M., et al. (2007), CRISM multispectral summary products: Harrison, K. P., and R. E. Grimm (2008), Multiple flooding events in the Parameterizing mineral diversity on Mars from reflectance, J. Geophys. Martian outflow channels, J. Geophys. Res., 113, E02002, doi:10.1029/ Res., 112, E08S14, doi:10.1029/2006JE002831. 2007JE002951. Phillips, R. J., et al. (2001), Ancient geodynamics and global‐scale hydrol- Harrison, K. P., and R. E. Grimm (2009), Regionally compartmented ogy on Mars, Science, 291, 2587–2591, doi:10.1126/science.1058701. groundwater flow on Mars, J. Geophys. Res., 114, E04004, Poulet, F., R. E. Arvidson, C. Gomez, R. V. Morris, J.‐P. Bibring, doi:10.1029/2008JE003300. Y. Langevin, B. Gondet, and J. L. Griffes (2008), Mineralogy of Terra Houston, J. (2006), Variability of precipitation in the Atacam Desert: Meridiani and western Arabia Terra from OMEGA/MEx and implica- Causes and hydrological impact, Int. J. Climatol., 26, 2181–2198, tions for their formation, Icarus, 195, 106–130, doi:10.1016/j.icarus. doi:10.1002/joc.1359. 2007.11.031. Hurowitz, J. A., W. W. Fischer, N. J. Tosca, and R. E. Milliken (2010), Putzig, N. E., M. T. Mellon, K. A. Kretke, and R. E. Arvidson (2005), Origin of acidic surface waters and the evolution of atmospheric chemistry Global thermal inertia and surface properties of Mars from the MGS on early Mars, Nat. Geosci., doi:2010/1038/ngeo831, in press. mapping mission, Icarus, 173,325–341, doi:10.1016/j.icarus.2004. Hynek, B. M. (2004), Implications for hydrologic processes on Mars from 08.017. extensive bedrock outcrops throughout Terra Meridiani, Nature, 431, Rossi, A. P., G. Neukum, M. Pondrelli, S. van Gesselt, T. Zegers, E. Hauber, 156–159, doi:10.1038/nature02902. A. Chicarro, and B. Foing (2008), Large‐scale spring deposits on Mars?, Hynek, B. M., and R. J. Phillips (2005), The etched terrain in Arabia Terra, J. Geophys. Res., 113, E08016, doi:10.1029/2007JE003062. Mars, is tilted, Lunar Planet. Sci., XXXVI, Abstract 1222. Saar, M. O., and M. Manga (2004), Depth dependence of permeability in Hynek, B. M., and R. J. Phillips (2008), The stratigraphy of Meridiani the Oregon Cascades inferred from hydrogeologic, thermal, seismic, Planum, Mars, and implications for the layered deposits’ origin, Earth and magmatic modeling constraints, J. Geophys. Res., 109, B04204, Planet. Sci. Lett., 274, 214–220, doi:10.1016/j.epsl.2008.07.025. doi:10.1029/2003JB002855. Hynek, B. M., R. E. Arvidson, and R. J. Phillips (2002), Geologic setting Schultz, P. H., and A. B. Lutz (1988), Polar wandering on Mars, Icarus, 73, and origin of the Terra Meridiani hematite deposit on Mars, J. Geophys. 91–141, doi:10.1016/0019-1035(88)90087-5. Res., 107(E10), 5088, doi:10.1029/2002JE001891. Sefton‐Nash, E., and D. C. Catling (2008), Hematitic concretions at Knauth, P. L., D. M. Burt, and K. H. Wohletz (2005), Impact origin of the Meridiani Planum, Mars: Their growth timescale and possible relation- sediments at the Opportunity landing site on Mars, Nature, 438,1123– ship with iron sulfates, Earth Planet. Sci. Lett., 269, 366–376, 1128, doi:10.1038/nature04383. doi:10.1016/j.epsl.2008.02.009. Kring, D. A. (2007), Guidebook to the geology of Barringer Meteorite Segura, T. L., O. B. Toon, A. Colaprete, and K. Zahnle (2002), Environ- Crater, Arizona, LPI Contrib., 1355. mental effects of large impacts on Mars, Science, 298, 1977–1980, Laskar, J., A. C. M. Correia, M. Gastineau, F. Joutel, B. Levrard, and doi:10.1126/science.1073586. P. Robutel (2004), Long term evolution and chaotic diffusion of the inso- Smith, D. E., et al. (2001), Mars Orbiter Laser Altimeter (MOLA): Experi- lation quantities of Mars, Icarus, 170, 343–364, doi:10.1016/j.icarus.2004. ment summary after the first year of global mapping of Mars, J. Geophys. 04.005. Res., 106, 23,689–23,722. Lewis, K. W., O. Aharonson, J. P. Grotzinger, R. L. Kirk, A. S. McEwen, Squyres, S. W., and J. F. Kasting (1994), Early Mars: How warm and how and T. ‐A. Suer (2008), Quasi‐periodic bedding in the sedimentary rock wet?, Science, 265, 744–749, doi:10.1126/science.265.5173.744. record of Mars, Science, 322, 1532–1535, doi:10.1126/science.1161870. Squyres, S. W., and A. H. Knoll (2005), Sedimentary rocks at Meridiani Lichtenberg, K. A., et al. (2010), Stratigraphy of hydrated sulfates in the Planum: Origin, diagenesis, and implications for life on Mars, Earth sedimentary deposits of Aram chaos, Mars, J. Geophys. Res., Planet. Sci. Lett., 240,1–10, doi:10.1016/j.epsl.2005.09.038. doi:10.1029/2009JE003353, in press. Squyres, S. W., et al. (2006), Planetary science: Bedrock formation at Malin, M. C., and K. S. Edgett (2000), Sedimentary rocks of early Mars, Meridiani Planum, Nature, 443,E1–E2. Science, 290, 1927–1936, doi:10.1126/science.290.5498.1927.

21 of 22 E06002 ANDREWS‐HANNA ET AL.: EARLY MARS HYDROLOGY E06002

Squyres, S. W., et al. (2009), Exploration of Victoria Crater by the Mars Weitz, C. M., M. D. Lane, M. Staid, and E. Z. Noe Dobrea (2008), Gray Rover Opportunity, Science, 324, 1058–1061, doi:10.1126/science. hematite distribution and formation in Ophir and Candor chasmata, 1170355. J. Geophys. Res., 113, E02016, doi:10.1029/2007JE002930. Stepinski, T. F., and A. P. Stepinski (2005), Morphology of drainage basins Williams, R. M. E. (2007), Global spatial distribution of raised curvilinear as an indicator of climate on early Mars, J. Geophys. Res., 110, E12S12, features on Mars, Lunar Planet. Sci., XXXVIII, Abstract 1338. doi:10.1029/2005JE002448. Wiseman, S. M., et al. (2008), Phyllosilicate and sulfate‐hematite deposits Tosca, N. J., S. M. McLennan, B. C. Clark, J. P. Grotzinger, J. A. Hurowitz, within Miyamoto crater in southern Sinus Meridiani, Geophys. Res. Lett., A. H. Knoll, C. Schroder, and S. W. Squyres (2005), Geochemical model- 35, L19204, doi:10.1029/2008GL035363. ing of the evaporation processes on Mars: Insight from the sedimentary Wiseman, S. M., R. E. Arvidson, R. V. Morris, F. Poulet, S. L. Murchie, record at Meridiani Planum, Earth Planet. Sci. Lett., 240, 122–148, F. P. Seelos, J. L. Bishop, J. C. Andrews‐Hanna, D. Des Marais, and doi:10.1016/j.epsl.2005.09.042. J. L. Griffes (2010), Spectral and stratigraphic mapping of hydrated Tosca, N. J., A. H. Knoll, and S. M. McLennan (2008), Water activity and sulfate and phyllosilicate‐bearing deposits in northern Sinus Meridiani, the challenge for life on early Mars, Science, 320, 1204–1207, Mars, J. Geophys. Res., doi:10.1029/2009JE003354, in press. doi:10.1126/science.1155432. Travis, B. J., N. D. Rosenberg, and J. N. Cuzzi (2003), On the role of J. C. Andrews‐Hanna, Department of Geophysics, Colorado School of widespread subsurface convection in bringing liquid water close to ’ Mines, 1500 Illinois St., Golden, CO 80401, USA. ([email protected]) Mars surface, J. Geophys. Res., 108(E4), 8040, doi:10.1029/ R. E. Arvidson and S. M. Wiseman, Department of Earth and Planetary 2002JE001877. ‐ Sciences, Washington University in St. Louis, Campus Box 1169, Wang, C. Y., M. Manga, and J. C. Hanna (2006), Can freezing cause 1 Brookings Dr., St. Louis, MO 63130, USA. floods on Mars, Geophys. Res. Lett., 33, L20202, doi:10.1029/ M. T. Zuber, Department of Earth, Atmospheric, and Planetary Sciences, 2006GL027471. Massachusetts Institute of Technology, Cambridge, MA 02139, USA.

22 of 22